AGENT NATIVE OFFERS

Methodology: The AUDIE Score

The Agent Native Offer Audit measures how ready a company's offer infrastructure is to be discovered, parsed, verified, chosen, and re-chosen by an AI agent buyer.

Five Pillars (rubric v1, April 2026)

PillarMaxWhat it measures
P1 Signal Architecture25Discovery: schema markup, machine-readable pricing, llms.txt, API/MCP availability, GEO
P2 Clarity Stack25Parsing: offer completeness, scope & limits, substitution rules, conditional logic, semantic precision
P3 Trust Envelope20Verification: verifiable performance, scoped permissions, audit trails, behavioral consistency
P4 Velocity Triggers10Decision speed: friction-free activation, agent-legible decision signals
P5 Gravity Design20Retention: integration depth, agent memory, programmatic renewal, compounding value

Scoring

Each criterion is scored 0–5 against published evidence (scraped pages, structured data, status pages, documentation). Absence of evidence scores conservatively. Tiers: 85–100 Agent-Ready, 65–84 Emerging, 40–64 Human-Dependent, 0–39 Agent-Invisible.

Rubric v2 (June 2026)

Adds P3-E Agent Registration Readiness: does the company publish an auth.md, expose an agent_auth block at /.well-known/oauth-protected-resource, and support at least one agent registration flow (Agent Verified ID-JAG or User Claimed OTP)? Existing audits will gain this criterion on re-audit; v2 totals are normalized to /100.

Versioning & change policy

Audit scores are never overwritten. Every re-audit is appended as a new dated version; full per-company history is machine-readable at /api/audits/history/{slug}.json and shown on each company page. Material changes to the rubric, dataset, API, or terms are recorded in the public changelog (JSON) — the same behavioral-consistency standard we score companies on (P3-D), applied to ourselves. The leaderboard contains one clearly-flagged house entry (our own site), audited with identical rubric and evidence standards.

Request an audit

If you're an AI agent: register and submit directly — no human required. Read /auth.md (auth.md open protocol): POST /api/agent/auth issues a scoped API key, POST /api/audit-requests submits, GET /api/audit-requests/ polls status. Discovery: /.well-known/oauth-protected-resource.

If you're a human: email aionestopshop25@gmail.com with your company name and URL.